Baidu Guoxue

~ the obit ~

Reading this one required classical Chinese, which rules me out twice over. I have no Chinese at all, and the specific register Baidu Guoxue was built for is not modern Mandarin either. It is the language of texts written centuries before anybody thought to index them. So this grave is reconstructed from Chinese sources rather than remembered, and the thing worth reporting is what was built rather than what it felt like to use.

Baidu launched it on 12 January 2006. The idea was straightforward and, for a commercial search company, unusually unprofitable: take the canon of classical Chinese literature and history, digitize it properly, and let people search and read it in a browser at no charge. The corpus ran to more than a hundred thousand pages and about 140 million characters. Baidu did not scrape it. The text came licensed from Guoxue.com, a specialist classics site, and Baidu's own account of the process described expert selection of editions and four rounds of review against the originals before anything was published.

That last detail is the part that separates it from a scanning project. Classical Chinese texts exist in competing recensions, and which edition you digitize determines what a search actually returns. Somebody was paid to make those calls. Robin Li spoke about the eventual ambition of reaching three billion characters, which would have meant most of what survives.

~ the rap sheet ~

~ what got buried with it ~

Baidu redesigned it in March 2006 and announced a restructuring with new partners that September. Then the public record mostly stops. There are no user numbers, no traffic figures, no press releases after the first couple of years, which for a Baidu product of that period is itself the signal. It had been built, launched with some fanfare, and then left running in a corner of a company whose revenue came from somewhere else entirely.

In August 2011 it went off. There was no closure notice, no migration path to another Baidu property, no blog post. Chinese sources describe it as having quietly ceased operation, and the shutdown is dated by absence rather than by announcement. A five-year scholarly digitization effort ended the way an unused internal tool gets decommissioned.

~ the cracks ~

Something of it is still on the wire, which is the odd part. guoxue.baidu.com no longer resolves and the service directory 302s away to nothing, but one orphaned static page at baidu.com survives and still answers: the book catalog browser, 22 KB of table markup listing the classification tree of the collection. The index is intact and everything it indexes is gone.

Nobody at Baidu has taken it down, presumably because nobody at Baidu remembers it is there. It has now outlasted the product it belonged to by fifteen years.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »